CN112508073A - Electric equipment infrared chart identification method combining deep learning with traditional algorithm - Google Patents
Electric equipment infrared chart identification method combining deep learning with traditional algorithm Download PDFInfo
- Publication number
- CN112508073A CN112508073A CN202011386688.3A CN202011386688A CN112508073A CN 112508073 A CN112508073 A CN 112508073A CN 202011386688 A CN202011386688 A CN 202011386688A CN 112508073 A CN112508073 A CN 112508073A
- Authority
- CN
- China
- Prior art keywords
- image
- deep learning
- network
- traditional algorithm
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000013135 deep learning Methods 0.000 title claims abstract description 19
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000001914 filtration Methods 0.000 claims abstract description 13
- 238000000605 extraction Methods 0.000 claims abstract description 5
- 238000012360 testing method Methods 0.000 claims description 12
- 238000012795 verification Methods 0.000 claims description 9
- 230000009466 transformation Effects 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 5
- 230000002708 enhancing effect Effects 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000007667 floating Methods 0.000 claims description 3
- 238000002156 mixing Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 238000007639 printing Methods 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims 1
- 238000001514 detection method Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000005855 radiation Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000003707 image sharpening Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/251—Fusion techniques of input or preprocessed data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Abstract
The invention discloses a method for identifying an infrared chart of power equipment by combining deep learning with a traditional algorithm, which comprises the following steps: performing smooth noise point operation on the data set by using Gaussian filtering; performing high-frequency information extraction operation on the data set by using a Laplace filter; using histogram transform assistance; and training the processed data set by using a fast RCNN network, adjusting parameters, and finally obtaining a model capable of effectively detecting the power equipment in the infrared chart. According to the method, effective characteristics of a data set are highlighted through a traditional algorithm, dependency among similar pixels is reduced, the characteristics are more obvious, and a model with higher accuracy can be trained effectively by a Faster RCNN network.
Description
Technical Field
The invention relates to a method for recognizing an infrared chart of electric power equipment by combining deep learning with a traditional algorithm.
Background
The infrared image of the power equipment is an infrared detection technology which detects infrared radiation energy emitted by the power equipment, converts the infrared radiation energy into corresponding electric signals and obtains a thermal image of the surface of the power equipment after the electric signals are processed; the infrared detection technology has the characteristics of long distance, no contact, no sampling, no disintegration, accuracy, rapidness, intuition and the like, is widely applied to detection and diagnosis of power equipment, and has important significance for improving the stability of a power system.
In the existing power transmission and transformation line, in order to ensure the normal operation of various power equipment in the power transmission and transformation line, the power equipment needs to be inspected at regular time, and the traditional manual inspection mode needs workers to judge the equipment state through a monitoring system, so that the problems of high labor cost, poor real-time performance, high misjudgment rate and the like exist; however, the intelligent inspection system for the electric power equipment using the traditional computer vision algorithm has many defects in processing the infrared heat map of the electric power equipment, and the identification rate of the electric power equipment is low. Meanwhile, in the existing intelligent power equipment inspection scheme based on deep learning, target detection algorithms (such as Faster R-CNN, YOLO and SSD) are used for processing infrared heatmaps, the algorithms firstly position power equipment in images and then perform identification operation, the identification efficiency is low, and the existing requirements cannot be met; therefore, a new method for identifying the infrared heat map of the power equipment needs to be invented to solve the problems encountered at present.
Disclosure of Invention
The invention aims to provide an electric power equipment infrared chart recognition method combining deep learning with a traditional algorithm, which highlights effective characteristics of a data set through the traditional algorithm, reduces the dependency among similar pixels, makes the characteristics more obvious, enables a fast RCNN network to effectively train a model with higher accuracy, and has practicability and wide application.
In order to solve the problems, the invention adopts the following technical scheme:
a method for recognizing an infrared heat map of electric equipment by combining deep learning with a traditional algorithm comprises the following steps:
1) acquiring infrared heat map data sets of 50 types of electric equipment, wherein each set is classified into 1000 pieces, extracting temperature information in the image, and storing the temperature information in floating point type data;
2) normalizing the data set, setting the size of the image to be 256 × 256, and distributing the values between 0 and 1;
3) performing histogram equalization operation on the image to improve the contrast;
4) filtering by using a Gaussian-Laplacian operator, enhancing details, ensuring that noise in original data is inhibited to a certain degree, and superposing a sharpened image with the original image;
5) randomly scrambling the sequence of the data set, and then randomly mixing the data set with the following data: 2: 2, dividing a training set, a verification set and a test set in proportion, and sending the training set, the verification set and the test set into a Faster RCNN network for training;
6) and adjusting network parameters to make the model reach the best.
Preferably, the step 2) comprises the step of carrying out linear transformation on the temperature data, and converting the processed data into decimal between (0, 1), so that the numerical sequence and the characteristics of the original data are not changed, and the training and convergence speed of the subsequent neural network can be accelerated.
Preferably, in step 3), the distribution of gray pixels in the image is changed through the histogram equalization process, the region with higher density of gray pixels is extended, and the range with lower density of gray pixels is narrowed, so that the gray levels in each image can be fully and effectively utilized, the image contrast is improved, and for a digital image with discrete gray levels, the probability is replaced by the frequency, and then the discrete form of the transformation function T (r { k }) can be expressed as:
wherein s { k } is the gray level of each pixel after equalization, and is also the gray level after normalization, the value is between 0 and 1, rk is the original pixel value, Pr is the probability function of rj in the original image, and nj is the number of pixels of which the pixel is j in the original image.
Preferably, the gaussian filtering in step 4) is a linear smoothing filtering, and is a weighted average of the whole image in the form of discretization window sliding window convolution, and the template coefficients of the whole image obey gaussian distribution.
Preferably, step 5) constructs a fast RCNN network, selects ResNet as its feature extraction part, preprocesses the input image through one convolutional layer and one pooling layer, then performs feature extraction through a residual network composed of 48 convolutional layers,
preferably, the input of the first convolution layer is 224 × 3, the longer side of the image is scaled by resize program, and finally 0 value is uniformly filled on both sides of the longer side, so that the shorter side reaches the standard length.
Preferably, the ResNet adopts a three-layer residual learning network, because 3 convolutional layers form one residual Block, the residual network of the ResNet is divided into four Block blocks, each Block respectively comprises 3, 4, 6 and 3 residual blocks, so that the four Block blocks are realized, and input parameters of each residual Block in the Block blocks are determined, wherein the input parameters comprise a third-layer output channel number depth, first-two-layer output channel numbers bottleck and a middle-layer step length stride.
Preferably, the ratio of 6: 2: 2, dividing the data set into a training set, a verification set and a test set, testing through a network, printing loss and acc information of the data set, analyzing a training result, adjusting parameters such as learning rate and the like, and finally generating an applicable model.
The invention has the beneficial effects that: the method has the advantages that effective characteristics of a data set are highlighted through a traditional algorithm, dependency among similar pixels is reduced, characteristics are more obvious, a Faster RCNN can effectively train a model with higher accuracy, meanwhile, through a preprocessing step of the data set, characteristic enhancement is carried out on sample data, the model is easier to converge in the training process, higher accuracy is obtained, and the method has practicability and wide application range.
Detailed Description
A method for recognizing an infrared heat map of electric equipment by combining deep learning with a traditional algorithm comprises the following steps:
1) acquiring infrared heat map data sets of 50 types of electric equipment, wherein each set is classified into 1000 pieces, extracting temperature information in the image, and storing the temperature information in floating point type data;
2) normalizing the data set, setting the size of the image to be 256 × 256, and distributing the values between 0 and 1;
3) performing histogram equalization operation on the image to improve the contrast;
4) filtering by using a Gaussian-Laplacian operator, enhancing details, ensuring that noise in original data is inhibited to a certain degree, and superposing a sharpened image with the original image;
5) randomly scrambling the sequence of the data set, and then randomly mixing the data set with the following data: 2: 2, dividing a training set, a verification set and a test set in proportion, and sending the training set, the verification set and the test set into a Faster RCNN network for training;
6) and adjusting network parameters to make the model reach the best.
Step 2) comprises the steps of carrying out linear transformation on the temperature data, converting the processed data into decimal between (0, 1) or (-1,1), enabling the numerical ordering and the characteristics of the original data not to be changed, accelerating the subsequent training and convergence speed of the neural network,
x'＝(x-X_min)/(X_max-X_min)；x'＝(x-X_mean)/σ。
and 3) changing the gray pixel distribution in the image through the histogram equalization processing, extending the area with higher gray pixel density, and reducing the range with lower gray pixel density, so that the gray levels in each image can be fully and effectively utilized, the image contrast is improved, and for the digital image with discrete gray levels, the frequency is used for replacing the probability, and then the discrete form of the transformation function T (r { k }) can be expressed as follows:
wherein s { k } is the gray level of each pixel after equalization, and is also the gray level after normalization, the value is between 0 and 1, rk is the original pixel value, Pr is the probability function of rj in the original image, and nj is the number of pixels of which the pixel is j in the original image.
Step 4), gaussian filtering is linear smooth filtering, and is to perform weighted average on the whole image in a discretization window sliding window convolution mode, wherein the template coefficient of the gaussian filtering is subjected to gaussian distribution, the gaussian distribution function refers to a probability density function of normal distribution, and the one-dimensional mode and the two-dimensional mode when the mean value mu is 0 are as follows:
the specific operation mode is that pixels in an original image are scanned one by using a preset convolution template, the weighted gray value of a target pixel neighborhood is calculated and determined by using the convolution template, and finally the value of a target pixel point is replaced.
The smooth filtering of the image is a suppression to the high-frequency gray level jump of the image, the sharpening of the image is opposite, the image is an enhancement to the high-frequency jump part of the image, information such as edge detail change of the image is highlighted, but the sharpening also enhances the noise of the image to a certain extent, so in the invention, two means are combined for suppressing the noise and enhancing the detail respectively, when the original image has obvious noise, the smooth filtering is firstly carried out and then the sharpening is carried out, if the sharpened image has the noise, then the noise reduction treatment can be further carried out, a laplacian operator is used in a sharpening filter, the laplacian operator is a differential operator, the main purpose is to enhance the part with the sudden change of gray level pixels in the image, and the area with the slow change of the gray level is weakened, and the operator is defined as:
in the discrete case, it can be approximated as:
V2f(x，y)＝f(x+1，y)+f(x-1，y)+f(x，y+1)+f(x，y-1)-4f(x，y)
it can be seen that the laplacian operator is a convolution template, and the absolute value of the difference between the central element value 4 times or 8 times and the field value thereof is calculated, and can also be understood as the absolute value of minus 5 times of the central element and the element and mean value in the field thereof; in the current sharpening process, firstly, a Laplacian filter is used for processing an image to generate an image highlighting the gray level mutation details, and then the Laplacian image and an original image are superposed to generate a sharpened image; therefore, the gray information in the original image can be reserved, and the interested detail parts in the image can be selectively highlighted; the method mainly uses the following two operators to process the image:
operator one [ [0,1,0],
[1,-4,1],
[0,1,0]]
the operator two [1,1,1],
[1,-8,1],
[1,1,1]]
the most middle value represents the peak, the closer the value is to 8, the better the sharpening effect is, the larger the value is, the image sharpening effect is gradually weakened, and the specific gravity of the central pixel is larger.
Step 5) constructing a fast RCNN network, selecting ResNet as a characteristic extraction part of the fast RCNN network, preprocessing an input image by the network through a convolution layer and a pooling layer, then extracting characteristics by a residual error network formed by 48 convolution layers, wherein the input of the first convolution layer is 224 x 3, zooming the longer side of the image by a resize program, and finally uniformly filling 0 values on two sides of the longer side to ensure that the shorter side reaches the standard length;
the temperature information of the infrared thermograph is taken as data, an image channel modification program is added, the input image is modified into three channel images by copying the data of a single channel, the data on the three channels of the image are the same, and the detection precision is not influenced;
ResNet adopts a three-layer residual error learning network, 3 convolutional layers form a residual error Block, the residual error network of ResNet is divided into four Block blocks, each Block Block respectively comprises 3, 4, 6 and 3 residual error blocks, the four Block blocks are realized, and input parameters of each residual error Block in the Block blocks are determined, wherein the input parameters comprise a third layer output channel number depth, a first two layers output channel number bottellen and a middle layer step size stride;
and (3) adding the following components in percentage by weight of 6: 2: 2, dividing the data set into a training set, a verification set and a test set, testing through a network, printing loss and acc information of the data set, analyzing a training result, adjusting parameters such as learning rate and the like, and finally generating an applicable model.
The method performs characteristic enhancement on the sample data through the preprocessing step of the data set, so that the model is easier to converge in the training process, and a higher accuracy is obtained.
With respect to deep learning: deep learning is a very popular application direction in the field of machine learning, and a plurality of research results are obtained in the aspects of computer vision, natural language processing, bioinformatics and the like, and especially the effect far exceeding that of a traditional algorithm is obtained in the field of image recognition. Deep learning is to train a large number of marked picture samples by constructing a deep convolutional neural network to obtain an algorithm model which can fit real image information, and to perform image recognition operation by using the model.
The above is only a specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that are not thought of through the inventive work should be included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope defined by the claims.
Claims (8)
1. A method for recognizing an infrared chart of electric equipment by combining deep learning with a traditional algorithm is characterized by comprising the following steps of: the method comprises the following steps:
1) acquiring infrared heat map data sets of 50 types of electric equipment, wherein each set is classified into 1000 pieces, extracting temperature information in the image, and storing the temperature information in floating point type data;
2) normalizing the data set, setting the size of the image to be 256 × 256, and distributing the values between 0 and 1;
3) performing histogram equalization operation on the image to improve the contrast;
4) filtering by using a Gaussian-Laplacian operator, enhancing details, ensuring that noise in original data is inhibited to a certain degree, and superposing a sharpened image with the original image;
5) randomly scrambling the sequence of the data set, and then randomly mixing the data set with the following data: 2: 2, dividing a training set, a verification set and a test set in proportion, and sending the training set, the verification set and the test set into a Faster RCNN network for training;
6) and adjusting network parameters to make the model reach the best.
2. The method for recognizing the infrared heat map of the electric power equipment by deep learning and combining with the traditional algorithm as claimed in claim 1, wherein the method comprises the following steps: and step 2) linear transformation is carried out on the temperature data, the data are converted into decimal between (0, 1) after being processed, numerical sequencing and characteristics of original data are not changed, and the training and convergence speed of a subsequent neural network can be accelerated.
3. The method for recognizing the infrared heat map of the electric power equipment by deep learning and combining with the traditional algorithm as claimed in claim 1, wherein the method comprises the following steps: and 3) changing the gray pixel distribution in the image through the histogram equalization processing, extending the area with higher gray pixel density, and reducing the range with lower gray pixel density, so that the gray levels in each image can be fully and effectively utilized, the image contrast is improved, and for the digital image with discrete gray levels, the frequency is used for replacing the probability, and then the discrete form of the transformation function T (r { k }) can be expressed as follows:
wherein s { k } is the gray level of each pixel after equalization, and is also the gray level after normalization, the value is between 0 and 1, rk is the original pixel value, Pr is the probability function of rj in the original image, and nj is the number of pixels of which the pixel is j in the original image.
4. The method for recognizing the infrared heat map of the electric power equipment by deep learning and combining with the traditional algorithm as claimed in claim 1, wherein the method comprises the following steps: and 4), Gaussian filtering is linear smooth filtering, the whole image is subjected to weighted average in a discretization window sliding window convolution mode, and the template coefficient of the whole image follows Gaussian distribution.
5. The method for recognizing the infrared heat map of the electric power equipment by deep learning and combining with the traditional algorithm as claimed in claim 1, wherein the method comprises the following steps: and 5) constructing a fast RCNN network, selecting ResNet as a characteristic extraction part of the fast RCNN network, preprocessing an input image by the network through a convolutional layer and a pooling layer, and then extracting characteristics by a residual error network consisting of 48 convolutional layers.
6. The method for recognizing the infrared heat map of the electric power equipment by deep learning and combining with the traditional algorithm as claimed in claim 5, wherein the method comprises the following steps: the input of the first convolution layer is 224 x 3, the longer side of the image is scaled by resize program, and finally 0 value is uniformly filled on both sides of the longer side, so that the shorter side reaches the standard length.
7. The method for recognizing the infrared heat map of the electric power equipment by deep learning and combining with the traditional algorithm as claimed in claim 6, wherein the method comprises the following steps: ResNet adopts the residual error learning network of three-layer, because 3 convolutional layers constitute a residual error Block, the residual error network of ResNet is divided into four Block blocks, each Block Block contains 3, 4, 6, 3 residual error blocks respectively, realizes four Block blocks, confirms the input parameter of each residual error Block in Block Block, including the output channel number depth of the third layer, the output channel number bottellen of the first two layers, the middle layer step size stride.
8. The method for recognizing the infrared heat map of the electric power equipment by deep learning and combining with the traditional algorithm as claimed in claim 7, wherein the method comprises the following steps: and (3) adding the following components in percentage by weight of 6: 2: 2, dividing the data set into a training set, a verification set and a test set, testing through a network, printing loss and acc information of the data set, analyzing a training result, adjusting parameters such as learning rate and the like, and finally generating an applicable model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011386688.3A CN112508073A (en) | 2020-12-01 | 2020-12-01 | Electric equipment infrared chart identification method combining deep learning with traditional algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011386688.3A CN112508073A (en) | 2020-12-01 | 2020-12-01 | Electric equipment infrared chart identification method combining deep learning with traditional algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112508073A true CN112508073A (en) | 2021-03-16 |
Family
ID=74969949
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011386688.3A Pending CN112508073A (en) | 2020-12-01 | 2020-12-01 | Electric equipment infrared chart identification method combining deep learning with traditional algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112508073A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115988181A (en) * | 2023-03-08 | 2023-04-18 | 四川三思德科技有限公司 | Personnel monitoring system and method based on infrared image algorithm |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107784661A (en) * | 2017-09-08 | 2018-03-09 | 上海电力学院 | Substation equipment infrared image classifying identification method based on region-growing method |
| CN108564565A (en) * | 2018-03-12 | 2018-09-21 | 华南理工大学 | A kind of power equipment infrared image multi-target orientation method based on deep learning |
| CN108898077A (en) * | 2018-06-15 | 2018-11-27 | 杭州天铂红外光电技术有限公司 | The recognition methods of power equipment infrared chart and power equipment infrared recognition system |
| CN110969610A (en) * | 2019-12-03 | 2020-04-07 | 杭州天铂云科光电科技有限公司 | Power equipment infrared chart identification method and system based on deep learning |
| CN111950566A (en) * | 2020-08-04 | 2020-11-17 | 国网安徽省电力有限公司电力科学研究院 | Rotation-invariant HOG infrared image power equipment identification method |
-
2020
- 2020-12-01 CN CN202011386688.3A patent/CN112508073A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107784661A (en) * | 2017-09-08 | 2018-03-09 | 上海电力学院 | Substation equipment infrared image classifying identification method based on region-growing method |
| CN108564565A (en) * | 2018-03-12 | 2018-09-21 | 华南理工大学 | A kind of power equipment infrared image multi-target orientation method based on deep learning |
| CN108898077A (en) * | 2018-06-15 | 2018-11-27 | 杭州天铂红外光电技术有限公司 | The recognition methods of power equipment infrared chart and power equipment infrared recognition system |
| CN110969610A (en) * | 2019-12-03 | 2020-04-07 | 杭州天铂云科光电科技有限公司 | Power equipment infrared chart identification method and system based on deep learning |
| CN111950566A (en) * | 2020-08-04 | 2020-11-17 | 国网安徽省电力有限公司电力科学研究院 | Rotation-invariant HOG infrared image power equipment identification method |
Non-Patent Citations (1)
| Title |
|---|
| 李浩: "基于计算机视觉的电力设备红外检测研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115988181A (en) * | 2023-03-08 | 2023-04-18 | 四川三思德科技有限公司 | Personnel monitoring system and method based on infrared image algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107564025B (en) | Electric power equipment infrared image semantic segmentation method based on deep neural network | |
| CN107578418B (en) | Indoor scene contour detection method fusing color and depth information | |
| WO2015180527A1 (en) | Image saliency detection method | |
| CN110084238B (en) | Finger vein image segmentation method and device based on LadderNet network and storage medium | |
| CN108805863B (en) | Method for detecting image change by combining deep convolutional neural network with morphology | |
| CN112233129A (en) | Deep learning-based parallel multi-scale attention mechanism semantic segmentation method and device | |
| CN110782385A (en) | Image watermark removing method based on deep learning | |
| CN113112498A (en) | Grape leaf scab identification method based on fine-grained countermeasure generation network | |
| CN102034105B (en) | Object contour detection method for complex scene | |
| CN112508073A (en) | Electric equipment infrared chart identification method combining deep learning with traditional algorithm | |
| Wiling | Locust Genetic Image Processing Classification Model-Based Brain Tumor Classification in MRI Images for Early Diagnosis | |
| CN109712095B (en) | Face beautifying method with rapid edge preservation | |
| CN109522859B (en) | Urban impervious layer extraction method based on multi-feature input of hyperspectral remote sensing image | |
| CN112967296B (en) | Point cloud dynamic region graph convolution method, classification method and segmentation method | |
| Umilizah et al. | Combination of Image Improvement on Segmentation Using a Convolutional Neural Network in Efforts to Detect Liver Disease | |
| Chen et al. | Fresh tea sprouts detection via image enhancement and fusion SSD | |
| CN111414788B (en) | Overlapped chromosome image segmentation method based on deformable U-shaped network | |
| CN113052234A (en) | Jade classification method based on image features and deep learning technology | |
| CN111830051A (en) | Transformer substation equipment oil leakage detection method and detection system based on deep learning | |
| Nawawi et al. | Comprehensive pineapple segmentation techniques with intelligent convolutional neural network | |
| CN112669260B (en) | Fundus image optic disc macula lutea detection method and device based on deep neural network | |
| CN117152746B (en) | Method for acquiring cervical cell classification parameters based on YOLOV5 network | |
| CN114155264B (en) | Footprint information extraction method based on double modules | |
| CN113723398B (en) | False film identification method for industrial radiographic negative film | |
| CN112633400B (en) | Shellfish classification and identification method and device based on computer vision |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210316 |
|
| RJ01 | Rejection of invention patent application after publication |